Splitting ShadowImage pairs

If you split a pair when there are write I/Os to the P-VOL, only some of the write I/Os may be written to the S-VOL. To ensure that all write I/Os are written to the S-VOL, stop I/Os to the P-VOL before splitting the pair. You can maintain data consistency of the S-VOL (maintain the order in which data is written to the S-VOL) by using a consistency group pair-split to split a pair, or by stopping all I/Os to the P-VOL before splitting the pair.

P-VOL data and S-VOL data are synchronized when the pair status changes from COPY(SP) or PSUS(SP) to PSUS. Because SI update copy is performed asynchronously, it takes some time before the pair status changes. To ensure data consistency in the P-VOL and the S-VOL after a split, stop write I/Os from the host to the P-VOL in advance. By this method, you can keep the P-VOL from being updated during a split, and ensure data consistency between the P-VOL and the S-VOL.

To split pairs quickly, stop host access to the P-VOL before splitting the pairs.

If you split a pair during initial copy, the initial copy operation is canceled, and it is performed again after the split. For initial copy, the maximum copying multiplicity (the number of jobs used for concurrent copy processing) for each pair is 24 for VSP One Block 20 series and 48 for VSP One Block 85, but the number might change to 1 if you specify the -split option for the paircreate command. As the result, copy processing might take longer than when the -split option is not specified. Additionally, if you enable the Quick/Steady Split Multiplexing local replica option, the maximum copying multiplicity is 48 even the when the -split option is specified.

The copy processing to split a pair requires some time to check if the differential data exists even if there is no differential data. The amount of time to check the differential data takes longer as the volume capacity increases, and it might take 10 minutes or more for 256 TB.